Source: Fortune
Summary
Andrew Ho, a former researcher at OpenAI, left the company after eight months to start his own business selling high-end reinforcement learning datasets to frontier AI labs. Ho believes that the labs are overvalued and that demand for inference will rise dramatically, but he’s “paranoid” about competition from cheap models forcing labs onto a faster treadmill of spending. Ho doesn’t buy the idea of recursive self-improvement (RSI) and thinks that research is limited by a lack of “research taste.” He’s starting a company to create datasets for tasks that models still can’t do.
Our Reading
The numbers tell one story.
Ho’s advice to his former colleagues: take the money while you can. The frontier labs are overvalued, and revenue may not catch up with spending. Ho’s new company will focus on creating datasets for tasks that models still can’t do, like long-horizon scientific reasoning and statistical analysis. Nvidia and Micron are the winners in this scenario, as they sell the chips to everyone in the race. Ho’s vested equity in OpenAI is a significant proportion of his net worth, and he’s stressing about what will happen to it.
Ho is now speaking in investor-brain, and his concerns are aligned with Wall Street’s fears that the companies funding the AI buildout won’t get their money back.
Author: Evan Null
